Dynamic Optimization of Solar Thermal Energy Systems with Storage Kody Powell Dr. John Hedengren Dr. Thomas Edgar TWCCC: September 2011
Overview of Solar Thermal Power Promising technology Systems approach insight into plant design 2
Solar Thermal vsphotovoltaic (PV) Energy Conversion Solar Thermal Sunlight Heat Mechanical Electricity Photovoltaic Cost ($/kwh) 0.12 1 (0.06 Projected) 2 0.18-0.23 1 Efficiency 3 ~18% ~12% Solar Irradiance Used Sunlight Electricity Direct Normal Irradiance (DNI) Global Horizontal (GHI) Scale Large Scale Large Scale & Distributed Storage Thermal Storage Battery Storage Dispatchableon a large scale Yes Impact on grid Small Large No 1 http://www.window.state.tx.us/specialrpt/energy/exec/solar.html 2 http://www.reuters.com/article/2009/08/24/us-energy-maghreb-desertec-sb-idustre57n01720090824?sp=true 3 http://solarbuzz.com/facts-and-figures/markets-growth/cost-competitiveness 3
WHAT STARTS HERE CHANGES THE WORLD The Potential of Solar Thermal Power Over 7000 MW more Announced in U.S. Kimberlina 5 MW Nevada Solar One 64 MW SEGS 354 MW Ivanpah 392 MW Sierra Sun Tower 5 MW Solana 280 MW Martin SEC 64 MW 4
Can Forecasts Help Solar Thermal? Forecasting technology advancing How do we take advantage? Image from: Marquez, R. and Coimbra, C. F. M., Forecasting of global and direct solar irradiance using stochastic learning methods, ground experiments and the NWS database, Solar Energy, Volume 85, 2011 Look at extreme scenarios Compare Standard Control Approach to Dynamic Optimization 5
Constant Temperature and Constant NMPC approach Measure DNI, use as FF to plant Constant temperature, constant power Relief pipe used when hot tank fills Power Control u( t) t= T t= 0 ( ) min φ x( τ ), y( τ ), u( τ ), d( τ ) dτ ( ( τ ), ( τ ), ( τ ), ( τ )) ( ) ( ( τ ), ( τ ), ( τ ), ( τ )) x& = f x y u d 0 = g x( τ ), y( τ ), u( τ ), d( τ ) τ = 0, = 0 h x y u d [ t t T ] 6
Constant T Constant Po Approach Performance Improved by: Optimal temperature Hybrid Operation Storage Bypass Further Improvements Needed: Strategy for more stable operation Consider stochastic problem
Dynamic Optimization w/ Forecasts Hypothesis: Performance can be improved by: Controlling to optimal temperatures Hybrid operation Ability to bypass storage More DOFs u( t) t= T t= 0 min Po( τ ) ( ) dτ 8
Dynamic Optimization w/ Forecast Performance Improved by: Optimal temperature Hybrid Operation Storage Bypass Further Improvements Needed: Strategy for more stable operation Consider stochastic problem
Dynamic Optimization Improves Performance Sunny Day Solar Energy Collected (MWh) Energy Collected/ Total Incident Energy (%) Standard Control 18.02 76.8% Dynamic Optimization 18.59 79.2% Partly Cloudy Day Standard Control 14.60 75.8% Dynamic Optimization 15.83 81.1% Mostly Cloudy Day Standard Control 4.75 52.1% Dynamic Optimization 7.80 85.4% 10
Proposed D-RTO Formulation Supervisory Control NMPC for stability (run every 2-5 minutes) D-RTO runs every 1-2 hours Fewer variables may help solver find global min Forecast and plant states updated regularly Disturbances Plant Inputs Weather Database Plant NMPCs Forecasted Disturbances Plant States D-RTO Optimal Setpoints 11
Opportunities for Further Improvement More DOFs more optimal solutions Consider entire plant Include electric grid and demand forecasts Economic optimization Apply to other systems 12
WHAT STARTS HERE CHANGES THE WORLD Expanding the Potential of Solar Thermal Power 13
Conclusions Better utilization of renewable resources Particularly on cloudy days Systems approach leads to design insights Hybridization can greatly expand solar thermal utilization Supervisory D-RTO needed Thanks to APMonitor, National Science Foundation, Cockrell School of Engineering, UT-Austin, Edgar Group 14